Skip to content

Commit 6a1efee

Browse files
committed
fix: disable grounding tools for Gemini 3 models
Gemini 3 tool calls (googleSearch, urlContext) cause timeout via LiteLLM. This was causing %ask to hang while %code worked (no tools passed). See: google/adk-python#3627
1 parent cbce981 commit 6a1efee

1 file changed

Lines changed: 3 additions & 2 deletions

File tree

‎plugins/llm/src/llm/service.py‎

Lines changed: 3 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -472,12 +472,13 @@ def _get_gemini_tools(self, model: str) -> list[dict[str, dict]] | None:
472472
"""
473473
model_lower = model.lower()
474474

475-
# Supported Gemini 2.0+ text models (explicit opt-in)
475+
# Supported Gemini text models for grounding tools (explicit opt-in)
476+
# NOTE: gemini-3-flash excluded - tool calls cause timeout via LiteLLM
477+
# See: https://github.com/google/adk-python/issues/3627
476478
supported_models = [
477479
"gemini-2.0-flash",
478480
"gemini-2.5-flash",
479481
"gemini-2.5-pro",
480-
"gemini-3-flash",
481482
"gemini-flash-latest", # Alias for latest flash model
482483
]
483484

0 commit comments

Comments
 (0)